PAF1 complex component Leo1 helps recruitDrosophila Myc to promotersJennifer M. Gerlacha, Michael Furrerb, Maria Gallanta, Dirk Birkela, Apoorva Baluapuria, Elmar Wolfa,and Peter Gallanta,c,1

aDepartment of Biochemistry and Molecular Biology, Biocenter, University of Würzburg, 97074 Würzburg, Germany; bZoological Institute, University ofZürich, 8057 Zürich, Switzerland; and cComprehensive Cancer Center Mainfranken, 97078 Würzburg, Germany

Edited by Robert N. Eisenman, Fred Hutchinson Cancer Research Center, Seattle, WA, and approved September 11, 2017 (received for review April 7, 2017)

The Myc oncogene is a transcription factor with a powerful grip oncellular growth and proliferation. The physical interaction of Mycwith the E-box DNA motif has been extensively characterized, but itis less clear whether this sequence-specific interaction is sufficient forMyc’s binding to its transcriptional targets. Here we identify thePAF1 complex, and specifically its component Leo1, as a factor thathelps recruit Myc to target genes. Since the PAF1 complex is typicallyassociated with active genes, this interaction with Leo1 contributesto Myc targeting to open promoters.

Drosophila | Myc | transcription | growth | PAF1

The Myc family of transcription factors (MYC, MYCN, andMYCL) are potent oncogenes thought to contribute to the

majority of human cancers (1, 2). Through their C-terminal basic-helix-loop-helix zipper region (bHLHZ), Myc proteins dimerizewith the bHLHZ protein Max and subsequently bind their targetgenes (3) at promoter-proximal sites or at distally located en-hancers (4–7). High-affinity Myc targets often contain E-boxes(CACGTG and variants thereof) that are directly contacted byMyc:Max dimers. Many of these genes are involved in translation,or ribosome biogenesis or function (7–10). When expressed atsupraphysiological levels, Myc also invades most active promoters(11–16), and Myc has been proposed to act as general amplifier oftranscription (11, 12). The mechanism by which Myc recognizes itsdifferent targets is currently under investigation.In vitro Myc:Max dimers most efficiently bind to E-box motifs,

and this motif is frequently found in high-affinity Myc targets (7,10). However, several lines of evidence suggest that primaryDNA sequence does not suffice to define Myc targets. First, invivo Myc does not bind to all E-boxes indiscriminately, but ratherpreferentially associates with sites located in active promoters (5,13). Second, Myc binds to numerous promoters lacking anycommon sequence motif, especially when expressed at supra-physiological levels (e.g., ref. 13). Third, the measured in vitrobinding constant of Myc:Max dimers for E-boxes does not sufficeto explain the observed in vivo affinity for its target sites (10).Thus, it has been proposed that additional proteins help recruitMyc in a sequence-independent manner to its target sites (13).One such factor is the recently discovered WDR5, which inter-acts specifically with the highly conserved centrally locatedMBIIIb motif in Myc (10, 17). Loss of the WDR5 interactiondomain strongly reduces the binding of Myc:Max complexes totheir targets, including sites with canonical E-boxes.However, WDR5 likely is not the only such factor, since Myc

mutants lacking the WDR5 interaction region still retain partialtransforming activity in tissue culture and in mice (17). In addition,the MBIIIb region is highly conserved in Drosophila Myc, but amutant Myc derivative lacking the entire region can substitute forwild-type Myc in null mutant flies (18). Additional recruitmentfactors might include various sequence-specific transcription factorswith which Myc has been reported to associate (notably Miz1, Sp1,NF-Y, Smad2/3, and YY1; reviewed in ref. 19) and several com-ponents of the core transcription machinery (15, 20–22), includingBrf, P-TEFb, TBP, TFII-I, and RNA polymerase II (23, 24–29).

We decided to search for molecular partners that contribute toMyc’s recruitment to DNA and/or mediate transactivation byMyc, using Drosophila as a model system. Fruit flies contain asingle Myc protein that primarily controls cellular and organis-mal growth (reviewed in ref. 30). Like its vertebrate homologs,Drosophila Myc functions by dimerizing with Max and binding totarget genes. Most of the directly Myc-activated genes controlribosome biogenesis and function. This is consistent with Myc’sbiological properties in flies and also with the described targetsof vertebrate Myc proteins expressed at physiological levels (e.g.,refs. 7 and 10). Illustrating this evolutionary conservation of Mycfunction, vertebrate MYC can substitute for Drosophila Myc inflies (31), and Drosophila Myc can partially replace c-Myc incultured murine fibroblasts (32).To identify proteins that influence Myc’s transcriptional output,

we carried out an RNAi screen in Drosophila S2 cells (33) andidentified the PAF1 complex as a functional interactor of Myc.The PAF1 complex is conserved from yeast to man. It consists ofthe core subunits Paf1 (atms in Drosophila), Cdc73 (also known ashyrax in Drosophila and Parafibromin in vertebrates), Leo1 (Atuin Drosophila), Ctr9, Rtf1 (34, 35), and, in vertebrate cells, themore loosely associated Ski8 (36). The complex was initiallyidentified as an RNA polymerase II-associated factor (e.g., refs. 37and 38) and subsequently shown to interact with some generalfactors involved in transcription elongation, including DSIF (35,39), SII (40), FACT (35), the “super elongation complex” SEC(41), P-TEFb, Cdk12, and Cyclin T (42). Consistent with theseobservations, the PAF1 complex has been shown to positivelyaffect RNA polymerase II pause release and transcriptionalelongation (42–44). Other studies have revealed an inhibitory ef-fect of the PAF1 complex on elongation (45–48). Thus, it has beenproposed that the genetic background and physiological state ofdifferent cells may affect the output of the PAF1 complex (42). Ineither case, depletion of the PAF1 complex affects transcript levelsof most genes, but to only a rather small degree; for example, Myc

Significance

We identify the PAF1 complex component Leo1 as a factor thathelps recruit Myc to its target genes. In particular when Myc isoverexpressed, Leo1 becomes limiting for transcriptional reg-ulation by Myc.

Author contributions: E.W. and P.G. designed research; J.M.G., M.F., M.G., D.B., A.B., and P.G.performed research; E.W. and P.G. analyzed data; and J.M.G. and P.G. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Published under the PNAS license.

Data deposition: The sequences reported in this paper have been deposited in the Ar-rayExpress archive, https://www.ebi.ac.uk/arrayexpress (accession nos. E-MTAB-5470,E-MTAB-5471, and E-MTAB-5472).1To whom correspondence should be addressed. Email: peter.gallant@biozentrum.uni-wuerzburg.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1705816114/-/DCSupplemental.

E9224–E9232 | PNAS | Published online October 16, 2017 www.pnas.org/cgi/doi/10.1073/pnas.1705816114

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

targets were shown to be up-regulated by 10% on knockdown ofCdc73 (48), and efficient depletion of Paf1 altered the average ex-pression of a group of 4,855 direct PAF1 targets by <10% (46).Nevertheless, the PAF1 complex is essential for the survival ofmetazoans. The contributions of individual complex components tothese biochemical activities have not yet been fully elucidated. Thus,Paf1 is essential for complex formation and for histone H3 lysine4 methylation, but Leo1 is dispensable for both (40, 49), and nospecific biochemical activity has been ascribed to this protein (44).Along with associating with general transcription factors, the

PAF1 complex also interacts with the sequence-specific tran-scription factors β-catenin (through Cdc73; ref. 50), STAT3(through Ctr9; ref. 51), p53 (40), Ci/Gli (through Cdc73; ref. 52),and GCN4 (39). These interactions have been proposed tocontribute to recruitment of the PAF1 complex to specific genes(as in the case of GCN4) or, conversely, to the recruitment of theinteracting transcription factor (as shown for STAT3). Myc wasalso recently shown to interact with Cdc73 and proposed to re-cruit the PAF1 complex and subsequently transfer it onto RNApolymerase II. Depletion of Cdc73 moderately increases theaverage expression of Myc targets (by ∼10%), suggesting that thePAF1 complex exerts a negative effect on Myc-dependent targetsin the examined setting (48).The extent to which the PAF1 complex promotes cellular pro-

liferation and contributes to transformation in vertebrates is un-clear. On one hand, the complex component Cdc73 was originallyidentified as a tumor suppressor (53), and the PAF1 complex hasbeen reported to reduce transcription and protein stability of theproto-oncogene Myc (54). On the other hand, posttranslationalmodification can convert Cdc73 to an oncogene (55), and over-expression of Paf1 promotes growth of NIH 3T3 cells and inducestumor formation in vivo (56).Here we show that the PAF1 complex functionally and physi-

cally interacts with Myc in Drosophila larvae and cells. Eliminationof several complex components impairs Myc-dependent growth invivo, reduces the recruitment of Myc to its target genes, and affectstheir expression in S2 cells and in vivo. This suggests that thePAF1 complex contributes to Myc’s binding to its target pro-moters, thereby promoting cellular growth and proliferation.

ResultsThe PAF1 Complex Contributes to Myc-Dependent Growth in Vivo. Toidentify novel transcriptional cofactors for Myc, we carried outan RNAi screen in S2 cells (a cell line derived from Drosophilaembryos). Out of 752 tested transcription-associated factors, wefound 33 proteins that consistently affected the activity of theMyc-dependent reporter, including the scaffold protein dHCF(33) and the components of the PAF1 complex. Depletion of thePAF1 proteins does not strongly affect Myc protein levels (Fig.S1A, and see below), but it does reduce the expression of variousMyc targets (see Fig. 6 and Fig. S1B), indicating that PAF1 actsas a positive cofactor for Myc-dependent transcription.To assess the importance of the PAF1 complex for Myc’s bi-

ological activities, we evaluated the consequences of PAF1 de-pletion in vivo. Myc primarily controls cellular and organismicgrowth in flies (30). A moderate reduction in Myc levels results insmall adult flies with disproportionally small bristles, indicatingthat Myc activity is particularly important for the growth of bristleprecursor cells (57, 58). Knockdown of the PAF1 proteins Rtf1,Paf1, and Leo1 in these cells phenocopies the Myc-mutant phe-notype and significantly reduces the size of adult bristles (Fig. 1A;the different magnitudes presumably reflect differences in knock-down efficiencies). This observation is consistent with participationof the PAF1 complex in Myc-dependent growth.To provide more direct evidence, we measured PAF1’s effect

on the areas of imaginal disk cell clones overexpressing Myc.Myc-overexpressing clones are significantly larger than controls,due to a strong increase in cell size (ref. 58 and Fig. 1B). De-

pletion of all tested PAF1 complex components strongly affectedthis overgrowth, but did not reduce the size of control clones(Fig. 1B). Simultaneous depletion of the PAF1 componentLeo1 slightly reduced the level of ectopically expressed Mycprotein (Fig. S1C); the reason for this is unclear, and no sucheffect of Leo1 depletion on endogenous Myc levels was ob-served), but had a clearly stronger impact on clonal area (Fig.1B). Strong effects on Myc-induced growth were also seen ondepletion of other PAF1 complex components. Thus, thePAF1 complex is limiting for the extra growth of imaginal diskcells that is driven by Myc overexpression.

Leo1 Physically Interacts with Myc to Recruit It to Individual Targets.To explore the molecular mechanism for these functional effectsof the PAF1 complex on Myc activity, we turned back to culturedcells. Since the PAF1 complex did not strongly affect Myc pro-tein levels, we considered the possibility that it might physicallyinteract with Myc and thereby influence its transcriptional out-put. We first investigated the PAF1 complex component Leo1(since a corresponding epitope-tagged version was available) andobserved a strong association with coexpressed Myc in S2 cells(Fig. 2A). This interaction did not require any of the previouslydescribed Myc domains, i.e., the Myc boxes 1 or 2 situated in thetransactivation domain, the Myc box 3 mediating binding toWDR5, or the C-terminally located Max-interacting bHLHZdomain. Instead, a broad central region of Myc was needed forthe binding to Leo1 (Fig. 2A). Within Leo1, a central regionbetween amino acids 442 and 474 was required for the associa-

Fig. 1. The Paf1 complex affects Myc activity in vivo. (A) The indicatedcomponents of the Paf1 complex were depleted in bristle precursor cells, andsizes of dissected scutellar bristles from adult males were measured.(B) Areas of wing imaginal disk clones (in pixels) with Myc overexpression(red bars) or without Myc overexpression (blue bars), on knockdown of theindicated PAF1 complex components. Knockdown of LacZ served as a controlin A and B. The numbers of bristles from independent flies (A) or of analyzedclones (B) are indicated in each bar. Error bars represent SEM. *P < 0.05,**P < 0.001, two-tailed Student’s t test, for comparison with the appropriateLacZ knockdowns.

Gerlach et al. PNAS | Published online October 16, 2017 | E9225

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

tion with Myc (Fig. 2B). No biochemical functions or specificstructures have been ascribed to this region.We next examined whether this Leo1–Myc interaction also oc-

curs at physiological levels of either protein. We failed to observecoimmunoprecipitation of the endogenous proteins (presumablybecause of the limitations of our anti-Leo1 antiserum), and thusestablished a stable S2 cell line that allows inducible expression ofHA epitope-tagged Leo1. In the absence of inducer, these cellsshowed low leaky expression of HA:Leo1; on addition of CuSO4,the levels of HA:Leo1 rose to levels approaching those of endog-enous Leo1 (Fig. 3C and Figs. S2A and S3A, 5–8). The addition ofdsRNA targeting Myc efficiently eliminated Myc from these cells(Figs. S2A and S3A, 9–12). Using these cells, we carried out aproximity-ligation assay (PLA; ref. 59) and found that HA:Leo1 associates with endogenous Myc in nuclei (Fig. S3). More-over, endogenous Myc could be coimmunoprecipitated with HA:Leo1 in this setup, indicating that the PLA signal reflects a physicalinteraction between the two proteins (Fig. 3A). This interactiontakes place on DNA, as shown by sequential chromatin immuno-precipitation (Re-ChIP) experiments; previously identified Myctargets were specifically recovered when chromatin from HA:Leo1-expressing cells was immunoprecipitated with an anti-HAserum, followed by immunoprecipitation with anti-Myc anti-bodies (Fig. 3B). No such enrichment was seen in naïve S2 cellsor for control DNA sequences (Fig. 3B). Fig. 3C shows the ex-pression levels of HA:Leo1 and endogenous Myc. Together,these experiments show that Leo1 and Myc physically associatein cells. This interaction is likely direct, since it also can be seenbetween in vitro synthesized 35S-labeled Leo1 and a GST:Mycfusion protein purified from Escherichia coli (Fig. 3D). Weconsider it likely that this Leo1–Myc interaction serves to asso-ciate the entire PAF1 complex with Myc. Indeed, epitope-taggedversions of the other PAF1 complex components Paf1, Ctr9, andCdc73 can also be coimmunoprecipitated with coexpressed Myc in

S2 cells (Fig. 3E). These interactions may be mediated by en-dogenous Leo1, but we cannot exclude the possibility that addi-tional PAF1 components besides Leo1 directly interact with Myc.This physical interaction suggests two possibilities for how the

PAF1 complex might contribute to Myc’s activities: as a coactivatoror as a corecruiter. According to one scenario, Myc first binds to itstarget genes (via specific DNA motifs as well as other chromatinfactors, such as WDR5) and subsequently recruits the PAF1complex that then promotes transcriptional elongation on thesetargets. In the second scenario, the PAF1 complex localizes to

Fig. 2. Physical interaction of overexpressed Myc and Leo1. The epitope-tagged proteins were transiently expressed in S2 cells (as indicated above theblots) and immunoprecipitated with the indicated antibodies. (A) None of theconserved regions in Myc (N terminus, Myc boxes 2 and 3; C terminus consistingof a bHLHZ region) is individually essential for interaction with Leo1. Instead, acentral region of Myc is involved in binding to Leo1. (B) Leo1’s central regionbetween amino acids 442 and 474 is essential for binding to Myc.

Fig. 3. Myc and Leo1 associate at physiological expression levels.(A) Coimmunoprecipitation of ectopically expressed HA:Leo1 with endoge-nous Myc in S2 cells. HA:Leo1 expression was induced to near-physiologicallevels (C) with CuSO4 for 24 h before cell lysis. Immunoprecipitates wereincubated with the indicated antibodies to detect endogenous Myc and HA:Leo1. (B) Re-ChIP experiments were performed from HA:Leo1-expressing(Left, HA:Leo1) or naïve S2 cells (Right, control). Bound chromatin from ananti-HA chromatin immunoprecipitate was eluted with HA peptides andincubated with either anti-Myc (blue bars) or IgG as a control (red bars).(C) Western blot displaying the successful induction of HA:Leo1 was probedwith antibodies against HA, endogenous Leo1, and endogenous Myc; α-tubulinserved as a loading control. (D) Leo1 was translated in vitro in a rabbit re-ticulocyte lysate in the presence of 35S-labeled methionine; 10% of the trans-lation mixture was then loaded directly on a gel (lane “input”), and theremainder was incubated with a GST:Myc fusion protein (amino acids 46–507)purified from E. coli, or with GST as a control. After SDS/PAGE, the gel was fixedand dried, and radiolabeled Leo1 was visualized by autoradiography. (E) Theindicated epitope-tagged PAF1 complex components were transiently coex-pressed with either HA-tagged Myc (Left) or Myc carrying a 9E10 tag (Right),followed by immunoprecipitations with anti-AU1 or anti-HA antibodies. As-terisks in the lower input panel denote the PAF1 complex proteins.

E9226 | www.pnas.org/cgi/doi/10.1073/pnas.1705816114 Gerlach et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

open promoters (notably via contacts with the general transcriptionmachinery) and thereby acts as one of the chromatin-associatedfactors that helps recruit Myc to its targets. Subsequently, Mycstimulates target gene expression through other coactivators and/orby relieving the negative effects of the PAF1 complex on tran-scriptional elongation. The first scenario predicts that recruitmentof Leo1 to Myc target genes is dependent on the presence of Myc.Therefore, we assessed the binding of Leo1 to Myc targets. Sinceour affinity-purified anti-Leo1 antibody is not suitable for ChIP, weused anti-HA antiserum to assess the binding of HA:Leo1 to Myctargets with or without Myc knockdown. Quantitative depletionof Myc (Fig. S2B) did not strongly reduce binding of HA:Leo1 toMyc targets, indicating that Myc contributes only moderately toLeo1 recruitment to target genes (Fig. 4A).To address the alternative scenario, we knocked down Leo1 and

determined the binding of endogenous Myc to the same targetsdescribed above. Although Leo1 was only partially eliminated(Fig. S2C), Myc binding to these genes was significantly reduced(Fig. 4B). Reduced Myc binding was also observed when any ofthe other PAF1 complex components were depleted, although theeffects were less prominent than after Leo1 knockdown (Fig. 4C;Fig. S2D shows depletion efficiencies). These data show thatLeo1 and most likely the entire PAF1 complex are required forefficient binding of Myc to certain target genes.To investigate the impact of the PAF1 complex on Myc binding

on a global scale, we carried out ChIPseq experiments, focusing onMyc’s direct interactor Leo1, which had also shown the strongesteffect on Myc binding. In naïve S2 cells, we found Myc to bind714 regions (Table S1), including 296 localized in promoter-proximal regions and 166 overlapping with enhancers (as definedexperimentally in ref. 60). As expected, Myc knockdown stronglyimpaired these interactions; as examples, the genes analyzed inFig. 4 are shown in Fig. 5A (Fig. S2E shows exemplary depletionefficiencies). Myc binding to promoter and enhancer sites wassimilarly reduced (Fig. 5 B and C). In contrast, Leo1 knockdowndid not affect the interaction of Myc with enhancer sites (98% ofcontrol; P = 0.7907) (Fig. 5 B and C), attesting to the comparabilityof the different ChIPseq samples, but it significantly reduced thebinding of Myc to promoter sites (to 82% of control; P < 0.0001).The difference between enhancer and promoter sites is not relatedto the differential presence of Leo1 at the two types of sites, sinceHA:Leo1 was chipped with comparable efficiency to promoter andenhancer sites (Fig. 5B; Fig. S2F shows protein levels). Many ofthese promoters contain E-boxes (115 out of 296), which are di-rectly bound by Myc:Max complexes. The association of Myc withsuch E-box–containing promoters is somewhat more strongly im-paired by Leo1 depletion compared with the association withpromoters lacking E-boxes (P = 0.055) (Fig. 6D).

Leo1 Is Required for Efficient Expression of Myc Targets. The fore-going data indicate a requirement for Leo1 for efficient re-cruitment of Myc to its target promoters, suggesting a role forMyc’s transcriptional output. Therefore, we investigated theimpact of Leo1 on the cellular transcriptome of S2 cells. Despitefairly efficient reduction of Leo1 protein levels (Fig. S2G), theeffects on global transcription were modest (Fig. S4A and TableS2). However, Leo1 knockdown significantly reduced the ex-pression of genes with E-boxes immediately downstream of theirtranscription start sites (the majority of these genes were pre-viously shown to be Myc targets; ref. 61), as well as of genes thatwere bound by Myc over an E-box situated in the promoter andwhose expression decreased on Myc knockdown (presumablydirect Myc targets; ref. 7), consistent with the idea that Leo1-mediated Myc binding to target promoters is required for the fulltransactivation of these genes by Myc (Fig. 6A and Fig. S4B).The magnitude of these effects was small, however, possiblybecause PAF1’s positive effect on Myc recruitment is partiallyoffset by its negative effects on transcription (48).

Consistent with such small effects on gene expression in naïveS2 cells, depletion of Leo1 (and other PAF1 complex compo-nents) in control clones in vivo had little effect on clonal size(Fig. 1B). In contrast, the clonal overgrowth induced by Mycoverexpression was strongly impaired by simultaneous depletionof PAF1 complex components, suggesting that effects on Myctarget gene expression become more pronounced in such a set-ting. Thus, we set out to investigate the impact of Leo1 depletionon the transcriptomes of Myc-overexpressing imaginal disks,using the same animals and analogous conditions as for theclonal analysis shown in Fig. 1B. In addition, we assessed theconsequences of loss of Max on Myc-dependent transcription

Fig. 4. Mutual influence of Myc and Leo1 on chromatin binding. (A) Anti-HA ChIP from cells expressing HA:Leo1 or naïve S2 cells, with and withoutMyc depletion. Cells were harvested and processed at 24 h after the additionof Myc-dsRNA and 125 μM CuSO4 to induce expression of HA:Leo1. Shownare averages and SEMs from two biologically independent experiments,normalized to the signal obtained for Uhg1 from HA:Leo1-expressing con-trol cells. (B) ChIP of control cells or cells incubated with Leo1-dsRNA for 48 h,obtained with rabbit anti-Myc or control rabbit IgGs. Shown are averagesand SEMs from two to three biologically independent experiments, nor-malized to the signal obtained for Uhg1 from control cells. (C) ChIP fromcontrol cells or cells incubated with the indicated dsRNA for 48 h and pro-cessed as described for B. Data are the average of two biologically in-dependent experiments for each knockdown (six for control ChIPs), eachwith SEM.

Gerlach et al. PNAS | Published online October 16, 2017 | E9227

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

(either alone or in combination with Leo1 down-regulation),since our data suggested potentially related roles for Leo1 andMax in Myc recruitment.In an otherwise wild-type background, Myc overexpression sig-

nificantly induced 977 genes and repressed 1,176 genes (Fig. 6B).The Myc-induced genes largely fall into the previously recognizedcategories of Myc targets (notably ribosome biogenesis, ribosome

structure, translation, and mitochondrial function; Table S3) andshow an excellent overlap with the Myc target genes identified inS2 cells (e.g., Fig. S5A). The number of Myc-repressed genes issurprisingly large; Table S3 presents a Gene Ontology (GO)analysis of these genes. We and others have previously identifiedfewer than 500 Myc-repressed genes in S2 cells or in vivo, andtypically the number of repressed genes is smaller than the numberof activated genes (7, 61–63). It also has been noted that the Myc-repressed transcriptomes are less concordant across different ex-perimental settings, suggesting a greater impact of uncontrollableexperimental conditions than for the Myc-activated transcriptomes(7). Nevertheless, previously reported lists of Myc-repressed genescoincide to some extent with our present list of Myc-repressedgenes (Fig. S5B).Depletion of Leo1 reduced the median induction of the 977

Myc activated genes from 176% to 146% and restored the ex-pression of the 1,175 Myc-repressed genes from 57% to 77% ofcontrol (Fig. 6D). In particular, the effect on gene repression (arelative relief of 47%) was greater than the slight reduction inMyc levels caused by Leo1 knockdown (Fig. S1C), indicating thatLeo1 is required for efficient binding and control of target genesby Myc. The lack of Max affected Myc-repressed genes to aslightly greater degree (restoring expression to 85% of control),and almost completely eliminated Myc-mediated activation (re-ducing expression to 107% of control). A combination of the twolesions did not suppress Myc-dependent transcriptional controlany further (Fig. 6D). Very similar effects were observed whenthis analysis was restricted to a narrowly defined set of putativedirect Myc targets, 221 induced genes and 25 repressed genesthat were also bound by Myc in promoter regions in S2 cells (Fig.6E); we did not include Myc-bound enhancers, since we couldnot unambiguously assign transcripts to these enhancers. The lackof Max completely eliminated Myc-dependent transactivation ofE-box–containing targets, but allowed for some transactivation ofE-box–lacking Myc targets (Fig. S5C). These data show that Mycretained substantial transactivation potential in a Max null-mutantbackground in which essentially no Max transcripts were present(ref. 28 and Fig. 6C), and that even though the Leo1 knockdowndid not completely remove Leo1 transcripts (Fig. 6C), its conse-quences on global Myc-dependent gene induction and repressionwere significant. These analyses demonstrate a clear requirementfor Leo1 (and, by inference, the entire PAF1 complex) for Myc-dependent gene regulation, particularly in conditions of supra-physiological Myc activity.

Genetic Interaction of Leo1 and Max in Vivo.As an additional test tophenotypically assess the effects of Leo1 and/or Max on Myc-dependent processes, we turned to a previously described in vivoassay for Myc activity. Myc overexpression in developing eyespromotes the growth of individual ommatidia and also increasesthe rate of apoptosis in this tissue (64). The resulting adult eyesare larger overall and contain larger ommatidia, but their ar-rangement is disturbed, causing a rough appearance. The apo-ptosis of pigment cells produces an overall lighter eye colorationwith numerous interspersed dark ommatidia (Fig. 7A). De-pletion of Max in this context abrogates Myc’s growth-promotingability but does not affect its proapoptotic effects, therebyproducing a small, very rough-looking eye (28). In contrast,Leo1 knockdown does not impair Myc-dependent overgrowth,but does reduce eye roughness (i.e., Myc’s proapoptotic activity).Codepletion of Max and Leo1 eliminated much of this roughnessand Myc-dependent eye overgrowth (Fig. 7; compare the Maxknockdown with or without Leo1 knockdown). Importantly, inthe absence of Myc overexpression, neither Leo1 nor Maxknockdown had a strong effect on eye size or morphology,demonstrating that all of the described effects are mediated byMyc. These observations illustrate the importance of bothLeo1 and Max for Myc-dependent transcriptional control, and

Fig. 5. Global effects of Leo1 or Myc depletion on Myc binding in S2 cells(ChIPseq). (A) ChIPseq traces of Myc over the genomic regions that were alsoanalyzed by manual ChIP in Fig. 5. Asterisks indicate the positions of canonicalE-boxes. Lengths of the vertical axes (i.e., maximal read numbers) are 800 forUhg1, 500 for Uhg2, and 200 for NACα. (B) Myc binding is significantly reducedon Leo1 depletion compared with naïve S2 cells for sites located in promoters(296 sites; average reduction of the median, 22%; P < 0.0001, Wilcoxon signed-rank test), but not for sites located in enhancers (166 sites; average reductionof the median, 2%; P = 0.7907). The difference between the effects on pro-moter and enhancer sites is highly significant (P < 0.0001, two-tailed t test),whereasMyc depletion affects both types of sites to the same extent (78% and79% reductions relative to the median of control cells, respectively; P = 0.2940,two tailed t test). (Lower) Relative read number of anti-HA ChIPs. Both pro-moter and enhancer regions are significantly enriched for reads in HA:Leo1-expressing cells comparedwith naïve S2 cells (P < 0.0001, Wilcoxon signed-ranktest), but the two regions do not significantly differ with respect to enrichment(7.9× and 6.6× median enrichment, respectively; P = 0.1895, two tailed t test).(C) Average distribution of reads over Myc-binding sites located in promotersor enhancers (labeled “P” and “E”, respectively) in control cells or on depletionof Myc or Leo1. Reads were counted over 50-nt windows for the indicatedconditions, followed by subtraction of input reads for the correspondingwindow. (D) Leo1 depletion affects Myc binding to enhancer sites to the sameextent regardless of whether or not these sites contain canonical E-boxes(median binding relative to naïve S2 cells, 99% and 97%, respectively; P =0.2234, two-tailed t test), whereas promoter sites containing canonical E-boxesare affected more strongly by Leo1 depletion compared with promoter siteslacking E-boxes (median binding relative to naïve S2 cells, 74% and 83%, re-spectively; P = 0.0564, two-tailed t test).

E9228 | www.pnas.org/cgi/doi/10.1073/pnas.1705816114 Gerlach et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

raise the possibility that these two proteins influence differentsets of Myc target genes. However, our RNAseq analysis did notidentify any apoptosis-related Myc targets that were differen-tially affected by Leo1 and Max, and thus we cannot molecularlyexplain these different phenotypes. It remains possible that theyare caused by small transcripts that were not investigated in ourtranscriptome studies (e.g., miRNAs, tRNAs), or that they in-volve nontranscriptional functions of Myc (Discussion).

DiscussionAlthough the Myc:Max complex binds to specific DNA se-quences in vitro, this interaction is not sufficiently strong to allowefficient interaction with these motifs when they are embeddedin chromatin context (10, 13). Additional proteins are needed forthe observed binding of Myc to its targets in vivo, such as WDR5(17) and the PAF1 complex identified here. We specifically showthat most components of the PAF1 complex can associate withMyc, and that Leo1 does so in vivo and in vitro, suggesting that itis Leo1 that mediates the interaction between PAF1 and Myc(possibly in addition to other complex components). By virtue ofits association with the general transcription machinery, thePAF1 complex is preferentially localized to active promoters andthus ideally placed to help recruit Myc to relevant binding sites.We assume that additional factors participate in the recruitmentof Myc to chromatin, and that the relative contributions of thesefactors vary in different cellular backgrounds. Nevertheless, thePAF1 complex is essential for full Myc binding to its endogenoustargets in S2 cells. It may appear surprising that knockdown ofPAF1 components does not affect the expression of Myc targetsto the same extent as Myc binding does; however, such mildconsequences of PAF1 depletion on cellular transcriptomes de-spite stronger effects on chromatin-associated proteins andchromatin marks have been consistently reported (e.g., refs. 46and 48). This can be rationalized by the combination of positive(e.g., recruitment of Myc) and negative (e.g., inhibition ofelongation; ref. 48) contributions of the PAF1 complex to geneexpression. As a result, the net effect on mature transcript levelsis dampened—overall, it may appear to be either positive ornegative (42).Depletion of PAF1 proteins has a substantial impact on the

extra tissue growth induced by high levels of Myc in vivo. This maybe caused in part by a moderate reduction of Myc protein levels(the reason for which is unknown), as well as by impaired re-cruitment of Myc to a large number of target genes. In addition,we cannot exclude the possibility that the PAF1 complex also af-fects transcription-independent processes. Although Myc is bestknown for its role in controlling transcription, it also affects othercellular processes independent of transcription, such as DNAreplication (27), mRNA cap methylation and translation (65), andα-tubulin acetylation (66). A role for the PAF1 complex in the firsttwo processes (which obviously contribute to cellular proliferationand/or growth) is conceivable, but this has not been addressedso far.In contrast to cells with high expression of Myc, the growth of

wild-type eyes or control wing imaginal disk clones is not reducedby depletion of the PAF1 complex. Indeed, ubiquitous depletionof the PAF1 complex throughout the animal allows most flies todevelop to the pharate adult stage, and some escapers evencomplete development and eclose as adults (SI Materials andMethods). This effect may be explained in part by insufficientknockdown efficiencies (the available null mutants in Paf1,Cdc73, or Ctr9 do show a stronger phenotype), but it also sug-gests that the PAF1 complex is less essential under normalgrowth conditions. On the other hand, in tissues undergoingrapid growth (notably imaginal disks experiencing Myc over-expression), depletion of Leo1 clearly reduces the ability of Mycto regulate targets and impedes the associated overgrowth. BothMyc-repressed and Myc-activated genes are affected, consistent

Fig. 6. Effects of Leo1 and Max on the expression of Myc targets (RNAseq).(A) Distribution of relative expression levels in S2 cells [kernel density of log2

(Leo1-KD/GFP-KD)] for different gene groups: all 9,123 expressed genes(black curve; median, 0.001838), 58 direct Myc targets (i.e., genes that aresignificantly down-regulated after Myc depletion and bound by Myc over anE-box situated in the promoter region; red curve; median, −0.02579), and170 genes with a downstream E-box (see text; blue curve; median,−0.03935). The latter two groups differ significantly from the former, withP = 0.00067 and 0.024, respectively (nonparametric Spearman correlation).(B) Dot plots showing log2 read numbers of the indicated imaginal diskgenotypes (averages of three biologically independent replicates) for 8,251genes expressed above a minimal threshold in our samples (Methods). Dotsmarked in red are significantly (P < 0.05), by >1.5-fold, deregulated by Mycoverexpression in the corresponding genotype. (C) Relative read numbersfor the Leo1 and Max transcripts in imaginal disks. Each bar is derived fromsix biologically independent RNAseq samples of the indicated genotype,where wt indicates Max+/+ and control corresponds to no Leo1 knockdown.Values for wt control are set to 100%. Error bars represent SEM. (D) Effect ofMyc overexpression in different genotypes. The x-axis shows the log2 of theratio (expression on Myc overexpression in genotype X/expression in theabsence of Myc overexpression, in genotype X), whereas the y-axis shows theeffect of Myc overexpression for control wing disks (i.e., Max wt and noLeo1 knockdown). Genes were sorted according to their relative Myc effectin control wing disks and pooled into bins of 40. (E) Fold expression changesin response to Myc overexpression in the indicated genotypes. Shown are221 and 25 genes that are bound by Myc in promoter regions in S2 cells andthat are significantly induced or repressed, respectively, by Myc over-expression in an otherwise wild-type background (a subset of the genesmarked in red in the upper left plot of B). All bars show median expressionratios, and all bars differ significantly from one another (P < 0.01, Mann–Whitney U test), except where indicated by “ns” (nonsignificant).

Gerlach et al. PNAS | Published online October 16, 2017 | E9229

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

with the idea that Myc recruitment is impaired in this situation.This suggests that supraphysiological levels of Myc saturate theavailable PAF1 complex and possibly also other “recruitmentfactors.” Thus, such settings, as are notably encountered in manyhuman tumors, might be particularly sensitive to inhibition ofPAF1 activity. Taken together, our observations emphasize theimportance of Leo1 for the biological activity of Myc, but they donot demonstrate that all PAF1 activities are mediated by Myc.Indeed, the PAF1 complex interacts with several transcriptionfactors besides Myc, and it is conceivable that some of thesefactors also contribute to the growth-related functions of PAF1.It remains to be seen whether Leo1 and Max are involved in

recruiting Myc to different functional sets of targets. The analysisof Myc-overexpressing adult eyes suggests that Leo1 and Maxpredominantly affect different Myc-dependent processes: apo-ptosis and growth, respectively. Indeed, individual genes aredifferentially affected by Max or Leo1 knockdown, but no genesets obviously marked as “growth-related” or “apoptosis-related”behave in the expected manner. It is conceivable that the gene(s)responsible for apoptosis in the eye specifically disrupt pupal eyedevelopment (a 4-d-long process that mostly involves cellulargrowth and differentiation, but little proliferation), and thusmight not be recognized as being generally apoptosis-related.Alternatively, the relevant genes code for small transcripts(e.g., miRNAs, tRNAs) that have not been included in ourtranscriptome analyses. Finally, we stress that even in the complete

absence of Max (28), Myc retains the ability to regulate a signifi-cant number of target genes. Since Myc is unlikely to bind to thesegenes on its own, this begs the question of what other partner cansubstitute for Max in such a situation.

Materials and MethodsFlies. The following flies were used in our analyses: sca-GAL4 (BloomingtonDrosophila Stock Center; 6479), UAS-Max-IR (line 2-7; ref. 28), UAS-GFP (E.Hafen), UAS-LacZ (B. Edgar), actin-FRT-CD2-FRT-GAL4 (K. Basler), and GMR-GAL4 3×(UAS-Myc) (characterized in ref. 28). Additional UAS lines for RNAiwere obtained from the Vienna Drosophila Resource Center: UAS-Rtf1-IR(27341), UAS-atms-IR (20876), and UAS-atu-IR (17490). Relevant genotypes forFig. 6 included hs-FLP actin5C-FRT-stop-FRT-GAL4 UAS-GFP UAS-p35 (wt ctr), hs-FLP actin5C-FRT-stop-FRT-GAL4 UAS-GFP UAS-Myc UAS-p35 (wt ctr Myc-overexpression), hs-FLP actin5C-FRT-stop-FRT-GAL4 UAS-GFP UAS-p35 UAS-Leo1-IR (wt Leo1-KD), hs-FLP actin5C-FRT-stop-FRT-GAL4 UAS-GFP UAS-MycUAS-p35 UAS-Leo1-IR (wt Leo1-KD Myc-overexpression), hs-FLP actin5C-FRT-stop-FRT-GAL4 UAS-p35 Max−/− (Max ctr), hs-FLP actin5C-FRT-stop-FRT-GAL4UAS-Myc UAS-p35 UAS-GFP Max−/− (Max ctr Myc-overexpression), hs-FLPactin5C-FRT-stop-FRT-GAL4 UAS-GFP UAS-p35 UAS-Leo1-KD Max−/− (MaxLeo1-KD), and hs-FLP actin5C-FRT-stop-FRT-GAL4 UAS-GFP UAS-Myc UAS-p35UAS-Leo1-KD Max−/− (Max Leo1-KD Myc-overexpression). Relevant genotypesfor Fig. 7 included GMR-GAL4 UAS-GFP UAS-LacZ, GMR-GAL4 UAS-GFP UAS-Leo1-IR, GMR-GAL4 UAS-Max-IR UAS-GFP, GMR-GAL4 UAS-Max-IR UAS-Leo1-IR, GMR-GAL4 3×(UAS-Myc) UAS-GFP UAS-LacZ, GMR-GAL4 3×(UAS-Myc) UAS-GFP UAS-Leo1-IR, GMR-GAL4 3×(UAS-Myc) UAS-Max-IR UAS-GFP, and GMR-GAL4 3×(UAS-Myc) UAS-Max-IR UAS-Leo1-IR.

In Vivo Analysis. UAS-RNAi-transgenes were targeted to bristle precursor cellsusing the sca-GAL4 driver. Adult scutella were then dissected andmounted onglass slides in glycerol. Pictures were taken using a 5× lens, and bristle sizewas determined in Adobe Photoshop as the total pixel count a bristle coversin a picture. Clones expressing GFP or Myc + GFP were induced and analyzed(7) at 48 h after clone induction in wandering larvae. For RNAseq of imaginaldisks, flies of the appropriate genotypes were raised under standard con-ditions at 25 °C. At the age of 53–66 h (139–143 h for the Max−/− genotypes),they were subjected to a 2-h heat shock at 37 °C to induce ubiquitous ex-pression of the GAL4-dependent transgenes. Then, 48 h later, wing imaginaldisks were dissected into Qiazol and immediately stored at −80 °C untilfurther processing.

To determine ommatidial size, flies were raised under noncrowding condi-tions. Adult males were collected at 1–7 d after eclosion and killed by freezing.Eyes were photographedwith a Zeiss Discovery V8 stereomicroscope fittedwitha 1.5× lens and processed with Axiovision Extended Focus software. For eachgenotype, the area of 20 centrally located ommatidia was measured from atleast seven eyes from independent individuals.

RNAi Screen and S2 Cell Culture. Culture and transfection of S2 cells and theRNAi screen for Myc cofactors have been described previously (7, 33).

Molecular Biology. dsRNA-mediated knockdowns, quantitative real-time PCR,and manual chromatin-immunoprecipitations were carried out as describedpreviously (7). For Primers see Table S4. Dual luciferase assays were con-ducted as described previously at 48–60 h after transfection of reporters anddsRNA (7).

Antibodies. Antibodies were mouse anti-Drosophila Myc (7), rabbit anti-Drosophila Myc (Santa Cruz Biotechnology), mouse anti–α-tubulin (Sigma-Aldrich), rabbit anti-HA (Abcam or ICL), rabbit anti-AU1 (Bethyl Laborato-ries), mouse anti-AU1 (Covance), rat anti-Cdc73, and rabbit anti-Rtf1 (giftsfrom J. Lis). A polyclonal rabbit antiserum was raised against the peptideRDKVESQVESAPKEC (amino acids 356–369 of Drosophila Leo1) and affinity-purified (New England Peptide or ImmunoGlobe).

Plasmids for Expression in S2 Cells. Wild-type Myc and mutant derivativeswere cloned in-frame with an N-terminal hemagglutinin (HA) tag in thevector pUASattB. Numbered deletions (created by site-directed mutagenesis)retain the indicated regions of the Myc protein, e.g., amino acids 403–717.Mutants lacking specific Myc domains have been described previously (18,28, 33). In brief, they carry the following modifications: ΔN-term lacks aminoacids 1–293, ΔMB2 has the amino acid “GP” instead of amino acids 68–84(Myc box 2), ΔMB3 has amino acid “F” instead of amino acids 405–422 (Mycbox 3), ΔC-term lacks amino acids 626–717 (C terminus: bHLHZ), and ΔZ lacksamino acids 676–717 (leucine zipper). Analogously, the Leo1 coding region

Fig. 7. Effect of Leo1 and Max depletion on the Myc overexpression phe-notype in the eye. (A) Photomicrographs of male adult eyes of the indicatedgenotypes. (B) Average ommatidial size of flies with the genotypes shown inA. The number of analyzed independent eyes per genotype are indicated.Error bars represent SEM. Significance of deviation from the correspondinggenotype without knockdown transgene according to Student’s two-tailedt test: *P < 0.05; **P < 0.001. All genotypes within one series contain thesame number of UAS transgenes, to rule out titration of GAL4 (relevantgenotypes are described in Materials and Methods).

E9230 | www.pnas.org/cgi/doi/10.1073/pnas.1705816114 Gerlach et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

was cloned behind an HA or AU1 epitope tag in pUASattB, and mutantderivatives were generated and numbered as for the Myc mutants.

To obtain stable expression of HA:Leo1 (wild-type), the correspondingcoding region was inserted under control of the metallothionein promoter inthe vector pMT181 carrying a puromycin resistance marker (a gift of M. TiebeandA. Telemann, Zentrum fürMolekulare Biologie der Universität Heidelberg).On puromycin selection of stable cell pools, HA:Leo1 expression was induced byincubation with 125 μM CuSO4 for 24 h.

In Vitro Interaction. The E. coli strain BL21 was transformed with constructscoding for GST or a GST:Myc (amino acids 46–507) fusion, and protein expressionwas induced by the addition of 1 mM isopropyl β-D-1-thiogalactopyranoside for3 h at 37 °C. On bacterial lysis, GST proteins were purified by incubation withglutathione Sepharose beads (Amersham Biosciences). Leo1 was expressed invitro in a coupled rabbit reticulocyte lysate in the presence of 35S-labeled me-thionine (TNT Kit; Promega), and incubated with the GST/GST:Myc bound toglutathione beads in GST Binding Buffer (200 mMNaCl, 1 mM EDTA, 1 mMDTT,0.5% Nonidet P-40, 10% glyerol, and 0.05% BSA) containing protease inhibitors(Roche). Bead-bound proteins were then analyzed by SDS/PAGE followed byautoradiography, as described previously (33); 10% of the in vitro translationmix was directly loaded on the gel and served as input control.

Western Blot and Immunoprecipitation Analyses. For transient expression inS2 cells, appropriate UAS plasmids were cotransfected with tubulin-GAL4,and cells were harvested at 24–48 h after transfection. Pelleted cells werewashed once with cold 1× PBS and then lysed on ice for 30 min in lysis buffer[150 mM NaCl, 50 mM Tris·HCl pH 8.0, 5 mM EDTA pH 8.0, 0.5% NonidetP-40, containing Protease Inhibitor Mixture tablets (Roche)]. Insoluble con-tents were precipitated by centrifuging for 15 min at 16,200 × g. The lysateswere precleared for 1 h at 4 °C with protein G Sepharose bead suspension(GE Healthcare) and 5% of the lysate was set aside as input control. Incu-bations with 0.2–1 μg of the primary antibodies were performed at 4 °C for3 h, followed by a 1-h precipitation of the epitope antibody complexes withprotein G Sepharose beads. Some immunoprecipitations were performedwith Dynabeads (Life Technologies GmbH) which were incubated with 1 μgof the primary antibody for 6–8 h at 4 °C. 10% of the lysate were set aside asinput control. Lysates were incubated over night at 4 °C with the antibodycoupled beads. For all immunoprecipitations, the immunoprecipitated ma-terial was washed three times for 5 min in lysis buffer on ice, SDS samplebuffer was added, and the samples were analyzed by SDS/PAGE and im-munoblotting as described previously (33).

For coimmunoprecipitations of endogenousMycwith HA:Leo1, cells with astably integrated MT-HA:Leo1 plasmid were induced with 125 μM CuSO4. At24 h later, 1.5 × 108 cells were harvested, washed once with cold 1× PBS,lysed in Hepes-EDTA-glycerol-Nonidet P-40 (HEGN) buffer with 140 mM KCl,and sonicated for 40 s at 20% amplitude (Digital Sonifier Cell Disruptor;Branson). Five percent of the lysate was set aside as input control. Dyna-beads (Life Technologies) were preincubated with 8 μg of rabbit anti-HA(Abcam) primary antibody or control rabbit IgG for 6–8 h at 4 °C, and celllysates were incubated overnight at 4 °C with the antibody-coupled beads.The immunoprecipitate was washed three times with HEGN buffer con-taining protease inhibitors, SDS sample buffer was added, and the sampleswere analyzed by SDS/PAGE and immunoblotting as described above.

Immunostaining in Drosophila S2 Cells. Drosophila S2 cells were plated on poly-L-lysine (Sigma-Aldrich)–coated coverslips and exposed to 125 μM CuSO4 and/orMyc-dsRNA (2 μg/106 cells) for 24 h, fixed with 4% paraformaldehyde, per-meabilized with 0.3% Triton-X 100 and then treated with blocking solution(10% goat serum, 2% BSA, and 5% sucrose in PBS) for 45 min after washing.Cells were incubated overnight at 4 °C with primary antibodies in blockingsolution [rabbit anti-HA (Santa Cruz Biotechnology), 1:500 and mouse anti-Myc,at 0.3 μg/mL], washed in TBS with 0.1% Tween-20, incubated for 1 h with the

secondary antibodies at room temperature, and washed again. Cells weremounted on glass slides using aqua-fluoromount (Sigma-Aldrich) and imagedwith a confocal microscope (Nikon Ti-Eclipse) with a 60× objective. Images wereprocessed with ImageJ 1.50h (67).

Re-ChIP. Cell fixation, lysis, and sonication were carried out as described pre-viously for manual ChIP (7). On cell lysis, 1% of the lysate was set aside as inputcontrol. Anti-HA magnetic beads (Thermo Fisher Scientific) were prepared bythree washes with 1× PBS containing BSA (5 mg/mL), and 60 μL was incubatedwith the chromatin overnight at 4 °C. Dynabeads (Thermo Fisher Scientific) forthe secondary ChIP were similarly washed and incubated with 3 μg of anti-Mycantibody or control rabbit IgGs overnight. The precipitates were washed asdescribed for manual ChIP and then eluted twice with 0.8 mg/mL Pierce HApeptides (Thermo Fisher Scientific) in 1× RIPA buffer for 15 min at 37 °C. Fivepercent of the combined eluates were set aside as input control; the remainderwas incubated with the antibody-coupled Dynabeads for 6 h at 4 °C. Washing,elution, and extraction of the precipitates, as well as analysis by quantitativereal-time PCR, were carried out as described for manual ChIP.

RNAseq, ChIPseq, and Bioinformatic Analysis. RNAseq, ChIPseq, and bioinformaticanalyses were carried out as described previously (7), with the following modi-fications. Antibodies for ChIPseq were rabbit anti-Myc (Santa Cruz Biotechnology)and rabbit anti-HA (Abcam). Sequencing was done on an Illumina NextSeq 500.For each ChIPseq condition, as well as input control, 7,847,000 reads weremapped onto the reference genome dm6 (bowtie 2.2.4). Peaks were called withmacs 1.4.0 and statistically analyzed with R and GraphPad Prism.

Promoter regions were defined as the ±100 nucleotides flanking the an-notated transcription start sites (FlyBase FB2015_4) for a total of 17,716 pro-moters. Enhancers (5,499 regions) were derived from ref. 60, with coordinatesadapted to the reference genome dm6. Myc-binding sites were defined asthose identified by MACS in naïve S2 cells that did not overlap backgroundsites (as called by anti-HA ChIP from naïve S2 cells) and did not have increasedread numbers on Myc depletion (714 sites); 166 of these were located in en-hancer regions (as defined by ref. 60) and did not overlap promoters, whereas296 overlapped promoters. For the analysis shown in Fig. 6, reads werecounted in 300-nt windows centered on the Myc-bound summits (as called byMACS) using the bedtools v2.17.0 suite, reads for the input sample over thesame window were subtracted, and read ratios were calculated relative toChIPs from naïve control cells. ChIPseq profiles for Fig. 6Awere generated withthe genome browser IGB, using a MACS output retaining all duplicate reads(parameter “keep-dup all”).

For RNAseq of S2 cells, RNA was isolated in biologically independent tripli-cates at 48 h after Leo1 or control depletion, depleted of rRNA using theRibominusKit (Invitrogen), andprocessed for sequencing to adepthof 6,757,000(non-rRNA) reads. For final analysis, 9,123 geneswere keptwithat least one readin each of the six samples and at least one read per million on average for eitherthe control or Leo1 knockdown condition. Statistical analysis was performedwith the Bioconductor tools in R and with GraphPad Prism. For RNA isolationfrom larvae, between 16 and 27 imaginal disks per sample were dissected intoQiazol (Qiagen). Further processing was done as described previously (7), exceptthat cDNAs were prepared with NEBNext Poly(A) mRNA Magnetic IsolationModule (New England BioLabs). For each of the 24 samples (eight differentgenotypes × three replicates), an average of 6.3 million mapped reads wereobtained and analyzed as described above.

ACKNOWLEDGMENTS. We thank André Kutschke and Reinhold Krug for tech-nical support; the Bloomington Drosophila Stock Center, the Vienna DrosophilaResource Center, and the fly community for various fly lines; K. Basler, M. Tiebe,and A. Teleman for plasmids; and J. Lis for antibodies. Funding for this projectwas provided by the Swiss National Science Foundation (Grant 3100A0-120458/1,to P.G.) and the German Research Foundation (Grants WO 2108/1-1, to E.W.; GA1553/1-1, to P.G.; and GA 1553/2-1, to P.G.).

1. Conacci-Sorrell M, McFerrin L, Eisenman RN (2014) An overview of MYC and its in-

teractome. Cold Spring Harb Perspect Med 4:a014357.2. Gabay M, Li Y, Felsher DW (2014) MYC activation is a hallmark of cancer initiation and

maintenance. Cold Spring Harb Perspect Med 4:a014241.3. Lüscher B, Vervoorts J (2012) Regulation of gene transcription by the oncoprotein

MYC. Gene 494:145–160.4. Fernandez PC, et al. (2003) Genomic targets of the human c-Myc protein. Genes Dev

17:1115–1129.5. Guccione E, et al. (2006) Myc-binding site recognition in the human genome is de-

termined by chromatin context. Nat Cell Biol 8:764–770.6. Perna D, et al. (2012) Genome-wide mapping of Myc binding and gene regulation in

serum-stimulated fibroblasts. Oncogene 31:1695–1709.

7. Herter EK, et al. (2015) snoRNAs are a novel class of biologically relevant Myc targets.

BMC Biol 13:25.8. Chandriani S, et al. (2009) A core MYC gene expression signature is prominent in basal-

like breast cancer but only partially overlaps the core serum response. PLoS One 4:e6693.9. Ji H, et al. (2011) Cell-type independent MYC target genes reveal a primordial sig-

nature involved in biomass accumulation. PLoS One 6:e26057.10. Lorenzin F, et al. (2016) Different promoter affinities account for specificity in MYC-

dependent gene regulation. eLife 5:e15161.11. Lin CY, et al. (2012) Transcriptional amplification in tumor cells with elevated c-Myc.

Cell 151:56–67.12. Nie Z, et al. (2012) c-Myc is a universal amplifier of expressed genes in lymphocytes

and embryonic stem cells. Cell 151:68–79.

Gerlach et al. PNAS | Published online October 16, 2017 | E9231

BIOCH

EMISTR

YPN

ASPL

US

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0

13. Guo J, et al. (2014) Sequence specificity incompletely defines the genome-wide oc-cupancy of Myc. Genome Biol 15:482.

14. Sabò A, et al. (2014) Selective transcriptional regulation by Myc in cellular growthcontrol and lymphomagenesis. Nature 511:488–492.

15. Rahl PB, et al. (2010) c-Myc regulates transcriptional pause release. Cell 141:432–445.16. Walz S, et al. (2014) Activation and repression by oncogenic MYC shape tumour-

specific gene expression profiles. Nature 511:483–487.17. Thomas LR, et al. (2015) Interaction with WDR5 promotes target gene recognition

and tumorigenesis by MYC. Mol Cell 58:440–452.18. Schwinkendorf D, Gallant P (2009) The conserved Myc box 2 and Myc box 3 regions

are important, but not essential, for Myc function in vivo. Gene 436:90–100.19. Sabò A, Amati B (2014) Genome recognition by MYC. Cold Spring Harb Perspect Med

4:a014191.20. Roy AL, Carruthers C, Gutjahr T, Roeder RG (1993) Direct role for Myc in transcription

initiation mediated by interactions with TFII-I. Nature 365:359–361.21. Hateboer G, et al. (1993) TATA-binding protein and the retinoblastoma gene product

bind to overlapping epitopes on c-Myc and adenovirus E1A protein. Proc Natl Acad SciUSA 90:8489–8493.

22. Koch HB, et al. (2007) Large-scale identification of c-MYC–associated proteins using acombined TAP/MudPIT approach. Cell Cycle 6:205–217.

23. Eberhardy SR, Farnham PJ (2001) c-Myc mediates activation of the cad promoter via apost-RNA polymerase II recruitment mechanism. J Biol Chem 276:48562–48571.

24. Eberhardy SR, Farnham PJ (2002) Myc recruits P-TEFb to mediate the final step in thetranscriptional activation of the cad promoter. J Biol Chem 277:40156–40162.

25. Gargano B, Amente S, Majello B, Lania L (2007) P-TEFb is a crucial co-factor for Myctransactivation. Cell Cycle 6:2031–2037.

26. Rahl PB, Young RA (2014) MYC and transcription elongation. Cold Spring HarbPerspect Med 4:a020990.

27. Gomez-Roman N, Grandori C, Eisenman RN, White RJ (2003) Direct activation of RNApolymerase III transcription by c-Myc. Nature 421:290–294.

28. Steiger D, Furrer M, Schwinkendorf D, Gallant P (2008) Max-independent functions ofMyc in Drosophila melanogaster. Nat Genet 40:1084–1091.

29. Marshall L, Rideout EJ, Grewal SS (2012) Nutrient/TOR-dependent regulation of RNApolymerase III controls tissue and organismal growth in Drosophila. EMBO J 31:1916–1930.

30. Gallant P (2013) Myc function in Drosophila. Cold Spring Harb Perspect Med 3:a014324.31. Benassayag C, et al. (2005) Human c-Myc isoforms differentially regulate cell growth

and apoptosis in Drosophila melanogaster. Mol Cell Biol 25:9897–9909.32. Trumpp A, et al. (2001) c-Myc regulates mammalian body size by controlling cell

number but not cell size. Nature 414:768–773.33. Furrer M, et al. (2010) DrosophilaMyc interacts with host cell factor (dHCF) to activate

transcription and control growth. J Biol Chem 285:39623–39636.34. Mueller CL, Jaehning JA (2002) Ctr9, Rtf1, and Leo1 are components of the Paf1/RNA

polymerase II complex. Mol Cell Biol 22:1971–1980.35. Squazzo SL, et al. (2002) The Paf1 complex physically and functionally associates with

transcription elongation factors in vivo. EMBO J 21:1764–1774.36. Zhu B, et al. (2005) The human PAF complex coordinates transcription with events

downstream of RNA synthesis. Genes Dev 19:1668–1673.37. Shi X, et al. (1996) Paf1p, an RNA polymerase II-associated factor in Saccharomyces

cerevisiae, may have both positive and negative roles in transcription. Mol Cell Biol16:669–676.

38. Phatnani HP, Jones JC, Greenleaf AL (2004) Expanding the functional repertoire ofCTD kinase I and RNA polymerase II: Novel phosphoCTD-associating proteins in theyeast proteome. Biochemistry 43:15702–15719.

39. Qiu H, Hu C, Wong CM, Hinnebusch AG (2006) The Spt4p subunit of yeast DSIFstimulates association of the Paf1 complex with elongating RNA polymerase II. MolCell Biol 26:3135–3148.

40. Kim J, Guermah M, Roeder RG (2010) The human PAF1 complex acts in chromatintranscription elongation both independently and cooperatively with SII/TFIIS. Cell140:491–503.

41. He N, et al. (2011) Human Polymerase-Associated Factor complex (PAFc) connects theSuper Elongation Complex (SEC) to RNA polymerase II on chromatin. Proc Natl AcadSci USA 108:E636–E645.

42. Yu M, et al. (2015) RNA polymerase II-associated factor 1 regulates the release andphosphorylation of paused RNA polymerase II. Science 350:1383–1386.

43. Stevens JR, et al. (2011) FACT, the Bur kinase pathway, and the histone co-repressorHirC have overlapping nucleosome-related roles in yeast transcription elongation.PLoS One 6:e25644.

44. Tomson BN, Arndt KM (2013) Themany roles of the conserved eukaryotic Paf1 complex inregulating transcription, histone modifications, and disease states. Biochim Biophys Acta1829:116–126.

45. Bai X, et al. (2010) TIF1gamma controls erythroid cell fate by regulating transcriptionelongation. Cell 142:133–143.

46. Chen FX, et al. (2015) PAF1, a molecular regulator of promoter-proximal pausing byRNA polymerase II. Cell 162:1003–1015.

47. Crisucci EM, Arndt KM (2012) Paf1 restricts Gcn4 occupancy and antisense transcrip-tion at the ARG1 promoter. Mol Cell Biol 32:1150–1163.

48. Jaenicke LA, et al. (2016) Ubiquitin-dependent turnover of MYC antagonizes MYC/PAF1C complex accumulation to drive transcriptional elongation. Mol Cell 61:54–67.

49. Tenney K, et al. (2006) Drosophila Rtf1 functions in histone methylation, gene ex-pression, and Notch signaling. Proc Natl Acad Sci USA 103:11970–11974.

50. Mosimann C, Hausmann G, Basler K (2006) Parafibromin/Hyrax activates Wnt/Wgtarget gene transcription by direct association with beta-catenin/Armadillo. Cell 125:327–341.

51. Youn MY, et al. (2007) hCTR9, a component of Paf1 complex, participates in the tran-scription of interleukin 6-responsive genes through regulation of STAT3-DNA interac-tions. J Biol Chem 282:34727–34734.

52. Mosimann C, Hausmann G, Basler K (2009) The role of Parafibromin/Hyrax as a nu-clear Gli/Ci-interacting protein in Hedgehog target gene control. Mech Dev 126:394–405.

53. Carpten JD, et al. (2002) HRPT2, encoding parafibromin, is mutated in hyperparathyroidism-jaw tumor syndrome. Nat Genet 32:676–680.

54. Lin L, Zhang JH, Panicker LM, Simonds WF (2008) The parafibromin tumor suppressorprotein inhibits cell proliferation by repression of the c-myc proto-oncogene. ProcNatl Acad Sci USA 105:17420–17425.

55. Takahashi A, et al. (2011) SHP2 tyrosine phosphatase converts parafibromin/Cdc73 from atumor suppressor to an oncogenic driver. Mol Cell 43:45–56.

56. Moniaux N, et al. (2006) The human homologue of the RNA polymerase II-associatedfactor 1 (hPaf1), localized on the 19q13 amplicon, is associated with tumorigenesis.Oncogene 25:3247–3257.

57. Gallant P, Shiio Y, Cheng PF, Parkhurst SM, Eisenman RN (1996) Myc and Max ho-mologs in Drosophila. Science 274:1523–1527.

58. Johnston LA, Prober DA, Edgar BA, Eisenman RN, Gallant P (1999) Drosophila mycregulates cellular growth during development. Cell 98:779–790.

59. Söderberg O, et al. (2008) Characterizing proteins and their interactions in cells andtissues using the in situ proximity ligation assay. Methods 45:227–232.

60. Arnold CD, et al. (2013) Genome-wide quantitative enhancer activity maps identifiedby STARR-seq. Science 339:1074–1077.

61. Hulf T, et al. (2005) Whole-genome analysis reveals a strong positional bias of con-served dMyc-dependent E-boxes. Mol Cell Biol 25:3401–3410.

62. Orian A, et al. (2003) Genomic binding by the Drosophila Myc, Max, Mad/Mnt tran-scription factor network. Genes Dev 17:1101–1114.

63. Bonke M, et al. (2013) Transcriptional networks controlling the cell cycle. G3 (Bethesda) 3:75–90.

64. Montero L, Müller N, Gallant P (2008) Induction of apoptosis by Drosophila Myc. Genesis46:104–111.

65. Cowling VH, Cole MD (2007) The Myc transactivation domain promotes globalphosphorylation of the RNA polymerase II carboxy-terminal domain independently ofdirect DNA binding. Mol Cell Biol 27:2059–2073.

66. Conacci-Sorrell M, Ngouenet C, Eisenman RN (2010) Myc-nick: A cytoplasmic cleavageproduct of Myc that promotes alpha-tubulin acetylation and cell differentiation. Cell142:480–493.

67. Schneider CA, Rasband WS, Eliceiri KW (2012) NIH Image to ImageJ: 25 years of imageanalysis. Nat Methods 9:671–675.

68. Prober DA, Edgar BA (2000) Ras1 promotes cellular growth in the Drosophila wing.Cell 100:435–446.

69. Bolte S, Cordelieres FP (2006) A guided tour into subcellular colocalization analysis inlight microscopy. J Microsc 224:213–232.

70. Wood A, Schneider J, Dover J, Johnston M, Shilatifard A (2005) The Bur1/Bur2 com-plex is required for histone H2B monoubiquitination by Rad6/Bre1 and histonemethylation by COMPASS. Mol Cell 20:589–599.

E9232 | www.pnas.org/cgi/doi/10.1073/pnas.1705816114 Gerlach et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 17

, 202

0